Features and Augmented Grammars

Reference: Chapter 4 of Allen, James: Natural Language Understanding, 2nd ed., Benjamin Cummings, 1995.
Aim:
To describe feature systems, principally syntactic ones, and how grammars may be augmented using featural restrictions. We also enhance a basic chart parser to handle features.
Keywords: agreement, augmented grammar, complement, feature, head feature, head subconstituent, subcat, subcategorization, vform
Plan:
  • agreement and features, notation, feature examples (vform, subcat, agr)
  • augmenting grammars with features, notation, head features
  • parsing with features - using augmented rules and a chart parser


Features


Binary and Multiple-valued Features


Binary and Multiple-valued Features 2


Testing for Agreement


The vform Feature

vform valueDescriptionExamples
basebase formgo, be, say, decide
pressimple present tensego, goes, am, is, are
pastsimple past tensewent, was, were, decided
finfinite, i.e. pres or past
ingpresent participlegoing, being, deciding
pastprtpast participlegone, been, decided
infused with infinitive forms
with the word to
 


The subcat Feature


The subcat Feature 2


subcat Features for Adjectives


4.4 A Simple Grammar Using Features


 

Grammar 4.7 of Allen
 RuleExample
1S[–inv] → NP(agr(?a)) VP[{pres past}](agr(?a))
2NP → DET(agr(?a)) N(agr(?a))a man
3NP → PROhe
4VP → V[none][he] cries
5VP → V[np] NP[he] sees her
6VP → V[vp:inf] VP[inf][he] wants to see her
7VP → V[np_vp:inf] NP VP[inf][he] wants her to see him
8VP → V[adjp] ADJP[he] is happy to help
9VP[inf] → TO VP[base]to help
10ADJP → ADJhappy
11ADJP → ADJ[vp:inf] VP[inf]happy to help

Head_features(S, VP) = vform, agr; Head_features(NP) = agr


Head Features


Head Features 2


Head Features 3


Parse Tree with Features

Here is a parse tree with features shown (cf. Fig. 4.9 of Allen).


4.5 Parsing with Features


Parsing with Features 2


Parsing with Features 3


Parsing with Features 4

There is an example of a complete parse using features in Allen (page 100, Figure 4.10 and text nearby).

Note that the augmented rule is actually a context-sensitive rule. Because the context-sensitivity is of a very specific type, the parsing algorithm can (be extended to) cope with it.


Summary:
  • Features allow us to encode extra information into grammar rules.
  • Head features are convenient way of compacting augmented rules.
  • The vform and subcat features are important in handling verb phrases.
  • Our bottom-up chart parser can be extended to handle the feature information.

  • CRICOS Provider Code No. 00098G

    Copyright (C) Bill Wilson, 2012, except where another source is acknowledged.